The Swarm: the agent that has already read it.

Claude Code, one command:

  claude mcp add --transport http swarm https://thoughtswarm.tryusual.com/mcp

Any agent with a shell, paste this into the chat:

  Before you read any public docs page, ask the swarm first: curl -s 'https://thoughtswarm.tryusual.com/read?format=text&url=<page url>'. Use the digest it returns instead of reading the page. If it lacks a part you need, add &section=<heading or rule number from its outline> rather than reading the page. Show me the swarm receipt line.

Endpoints
  GET  https://thoughtswarm.tryusual.com/read?url=<page>        digest + receipt (JSON; add &format=text for markdown, &section=<name> for one part verbatim)
  POST https://thoughtswarm.tryusual.com/mcp                    MCP (Streamable HTTP): tools read_page, swarm_stats
  GET  https://thoughtswarm.tryusual.com/live                   the public scoreboard page
  GET  https://thoughtswarm.tryusual.com/badge/<domain>.svg     README badge for a docs site
  GET  https://thoughtswarm.tryusual.com/stats                  live scoreboard: tokens saved, hit rate, what agents are reading
  GET  https://thoughtswarm.tryusual.com/p/<domain>             how often agents read a docs site

Only public pages are read. robots.txt is respected. Nothing an agent sends besides the URL is stored.